An image-based multi-level modeling method is proposed to estimate the effective bulk thermal conductivity of lightweight aggregate (LWA) impregnated with phase change material (PCM). The pore structure heterogeneity of the LWA and the uneven distribution of the PCM are considered by the model. The microstructure of the LWA is represented at two length scales (micro-level and macro-level), which are analyzed individually using imaging techniques. The reconstruction process is conducted on two-dimensional images to obtain three-dimensional digital representative volume element of the LWA, upon which multi-level numerical modeling of heat conduction is performed to estimate the effective bulk thermal conductivity of the PCM-LWA. The computed thermal conductivity of 0.34 W/(m∙K) is closed to experimental result and compared with estimates based on five traditional analytical methods for the sake of superiority analysis. Finally, the simulated thermal conductivity is used in a concrete level simulation to show its feasibility for modeling the thermal behaviors of PCM-LWA-incorporated concrete mixtures, in which PCM are employed to mitigate the hydration heat and prevent thermal cracking in mass concrete. The scale-up simulation manifests that the PCM-LWA not only reduced the peak temperature in the concrete by 3.18 °C but also changed the temperature distribution and gradient pattern inside the concrete.